search

Silver-Shen / Causally-Regularized-Learning

A method which takes advantage of causal features for classification
24 stars 10 forks source link

hello,the Python version of causally-Regularized-learning #1

Open xiaowei-ui opened 5 years ago

xiaowei-ui commented 5 years ago

hello,dear author, do you have the python version of this algorithm? If so, can you tell me? Many thanks！looking forward to your reply...

Silver-Shen commented 5 years ago

@xiaowei-ui Working on it, python implementation will be released this fall.

xiaowei-ui commented 5 years ago

@ xiaowei-ui正在研究它，今年秋天将发布python实现。 that's ok,thank you!

xiaowei-ui commented 5 years ago

@ xiaowei-ui正在研究它，今年秋天将发布python实现。 that's ok,thank you!

@xiaowei-ui Working on it, python implementation will be released this fall.

excuse me，how is the prediction matrix I generated by the feature matrix X? The demo is through the ’double‘ function, what function should be used in the neural network?

Silver-Shen commented 5 years ago

@ xiaowei-ui正在研究它，今年秋天将发布python实现。 that's ok,thank you!

@xiaowei-ui Working on it, python implementation will be released this fall.

excuse me，how is the prediction matrix I generated by the feature matrix X? The demo is through the ’double‘ function, what function should be used in the neural network?

Which matrix I do you mean? There is no I in demo...

xiaowei-ui commented 5 years ago

@ xiaowei-ui正在研究它，今年秋天将发布python实现。 that's ok,thank you!

@xiaowei-ui Working on it, python implementation will be released this fall.

excuse me，how is the prediction matrix I generated by the feature matrix X? The demo is through the ’double‘ function, what function should be used in the neural network?

Which matrix I do you mean? There is no I in demo...

the sevnth line code of the balance_cost function,

Silver-Shen commented 5 years ago

@xiaowei-ui It is actually the indicator of whether a sample receive treatment, for binary feature, I = X_j > 0 (double only act as type conversion). We do not discuss continuous feature in our scope, but you can refer to causal inference literature for more information.

xiaowei-ui commented 4 years ago

@xiaowei-ui It is actually the indicator of whether a sample receive treatment, for binary feature, I = X_j > 0 (double only act as type conversion). We do not discuss continuous feature in our scope, but you can refer to causal inference literature for more information.

First of all, thanks for your reply,dear author,then if I use the pretrain model to extract features, what should I do with the features specificly to get the Xtrain(n*p) form in the paper, and how to handle the features to fit the input X matrix in the code?

Silver-Shen commented 4 years ago

@xiaowei-ui perform binarization on X would be enough i think